Lucace/mskcc-lung-adenocarcinoma-spatial-cohort

MSKCC Lung Adenocarcinoma Spatial & Clinical Survival Cohort

Curated, multi-center retrospective clinical and histological cohort of resected lung adenocarcinoma (LUAD) patients, designed for biophysical transport modeling, spatial microvascular decoupling, and clinical recurrence prediction in EGFR-mutant disease under targeted tyrosine kinase inhibitor (TKI) therapy (Osimertinib / ADAURA trial setting).

Dataset Summary

• Primary Cohort: Memorial Sloan Kettering Cancer Center (MSKCC 2020) cohort of $N = 604$ authentic resected lung adenocarcinoma patients.
• Validation Cohorts: TCGA LUAD ($N = 437$) and GSE31210 ($N = 226$) multi-center validation cohorts.
• Key Features:
  • Predominant histological architecture under IASLC consensus: Lepidic, Acinar, Papillary, Micropapillary, Solid.
  • Pathological staging (Stage I, II, III).
  • EGFR mutation status (Exon 19 del / L858R).
  • Clinical endpoints: Recurrence-Free Survival (RFS in months), Recurrence Event (0/1), Overall Survival (OS in months), Overall Survival Event (0/1).
  • Clinical covariates: Sex, Smoking history.

Cohort Subsets (Configs)

Use the dropdown above or specify the config name in Python:

Subset Config	Patients ($N$)	File	Description
default / mskcc_2020	604	mskcc_2020_real_cohort.csv	Primary curated MSKCC 2020 clinical cohort with IASLC histology, staging, and RFS/OS outcomes.
combined_massive	663	combined_filtered_massive_cohort.csv	Unified cross-validation cohort combining multi-center clinical datasets.
tcga_luad	437	tcga_luad_filtered_cohort.csv	Independent TCGA Lung Adenocarcinoma validation cohort.
gse31210	226	gse31210_filtered_cohort.csv	Independent microarray gene expression and clinical validation cohort.

Quickstart in Python

Using pandas:

import pandas as pd

# Load the primary MSKCC 2020 cohort (N = 604)
url = "https://huggingface.co/datasets/Lucace/mskcc-lung-adenocarcinoma-spatial-cohort/raw/main/mskcc_2020_real_cohort.csv"
df = pd.read_csv(url)

print(f"Total patients: {len(df)}")
print(df['PREDOMINANT_HISTOLOGIC_SUBTYPE'].value_counts())

Using Hugging Face datasets:

from datasets import load_dataset

# Load primary MSKCC 2020 cohort
dataset = load_dataset("Lucace/mskcc-lung-adenocarcinoma-spatial-cohort", "mskcc_2020")
df = dataset['train'].to_pandas()

# Load TCGA validation cohort
tcga_ds = load_dataset("Lucace/mskcc-lung-adenocarcinoma-spatial-cohort", "tcga_luad")
df_tcga = tcga_ds['train'].to_pandas()

Clinical & Biophysical Background

In adjuvant osimertinib therapy for resected EGFR-mutant NSCLC (ADAURA trial), approximately 68% of disease recurrences occur following scheduled 3-year treatment cessation, disproportionately concentrated in months 36–48. The vast majority of these late post-cessation relapses do not harbor secondary genetic gatekeeper mutations (such as C797S or MET amplification) and retain sensitivity to osimertinib rechallenge.

This cohort provides the empirical clinical foundation demonstrating that IASLC histological growth patterns govern drug penetration boundaries:

• Lepidic (low-density, alveolar preservation): Low recurrence rate (4.55%, HR = 1.00 reference).
• Solid (dense sheets, vascular compression): High recurrence rate (30.88%, HR = 6.75, Log-Rank $p = 6.33 \times 10^{-6}$).

Citation & Archival Reference

If you utilize this dataset in your computational oncology, mathematical biology, or digital pathology research, please cite:

@article{celebrano2026osimertinib,
  author       = {Luca Celebrano},
  title        = {Biophysical Reaction-Diffusion Transport and Evolutionary Sanctuary Dynamics in Compact EGFR-Mutant Lung Adenocarcinoma},
  year         = {2026},
  publisher    = {Zenodo},
  doi          = {10.5281/zenodo.23623653},
  url          = {https://doi.org/10.5281/zenodo.23623653}
}

Institutional Affiliation

Department of Clinical Medicine and Surgery, School of Medicine, Università degli Studi di Napoli Federico II, Naples, Italy.